Journal of Archaeological Science: Reports 17 (2018) 195–211 Contents lists available at ScienceDirect Journal of Archaeological Science: Reports journal homepage: www.elsevier.com/locate/jasrep Evidence for the breakdown of an Angkorian hydraulic system, and its MARK historical implications for understanding the Khmer Empire ⁎ Terry Lustiga, ,1, Sarah Klassenb, Damian Evansc, Robert Frenchd, Ian Moffate a Department of Archaeology, University of Sydney, NSW 2006, Australia b School of Human Evolution and Social Change, Arizona State University, 900 S Cady Mall, Tempe, AZ 85281, United States c École française d'Extrême-Orient, 22 Avenue du Président Wilson, Paris 75116, France d Independent Scholar, 30 Woodlands Rd, Ashbury, NSW 2193, Australia e Department of Archaeology, Flinders University, Bedford Park, SA 5042, Australia ARTICLE INFO ABSTRACT Keywords: This paper examines the construction and design of a 7-km long embankment, probably built for King Angkor Jayavarman IV between 928 and 941 CE, as part of a new capital. We calculate that the capacities of the outlets Archaeology were too small, and conclude that the embankment failed, probably within a decade of construction, so that the Engineering practice benefits of the reservoir stored by the embankment and the access road on top of it were lessened substantially. Koh Ker We explain how the design was sub-optimal for construction, and that while the layout had a high aesthetic Lidar impact, the processes for ensuring structural integrity were poor. Simple and inexpensive steps to secure the weir Palaeohydrology Trial and error were not undertaken. We speculate that this early failure may have contributed to the decision to return the Water management royal court and the capital of the Khmer Empire to the Angkor region, marking a critically important juncture in regional history. 1. Introduction there no later than 921 CE (Cœdès, 1931, 13; Cœdès, 1937, 50), while Harśovarman I (910–925 CE) and Iśanavarman I (925–928 CE), the With the Angkorian state having lasted for more than six centuries sons of Yaśovarman I, his uncle by marriage, were still enthroned at (9th to 15th centuries CE), scholars have long sought to understand Angkor (Cœdès, 1953, 98, 147). The first inscriptions to attest to what contributed to its sustainability and what led to its eventual de- Jayavarman's power over the Khmer realm do not appear until 928 CE cline (Evans et al., 2007; Fletcher et al., 2008; Groslier, 1979; Groslier, (Cœdès, 1931,13–16). Despite much debate on the topic (e.g., Cœdès, 2007). We suggest that some insights might be gained from studying 1931, 16; Jacques, 1971, 169), it is now generally recognized that what happened at Koh Ker, another Khmer political centre about 80 km Jayavarman was likely a legitimate heir to the throne, not a usurper ENE of Angkor (Fig. 1). There is mounting evidence from archae- (Vickery, 1986, 108).2 While it is yet to be agreed why the political ological excavation and survey for a long and complex history of oc- center shifted to Koh Ker, for this paper, we are seeking factors that cupation at Koh Ker (Evans, 2010–2011), but it is clear that the city was might help explain why its time there was so short. very short-lived as the centre of Khmer power, lasting only about It is clear from the inscriptions (Cœdès, 1937, 68) that Jayavarman 17 years as the capital, from 928 to 944 CE. constructed Prasat Thom during this period, its pyramid being the tal- Jayavarman IV, the first recorded ruler at Koh Ker, was established lest in the Khmer world at the time (Cœdès, 1937, 70). As well, just to Abbreviations: APHRODITE, Asian Precipitation – Highly Resolved Observational Data Integration Towards Evaluation (of Water Resources); ARI, annual recurrence interval; ASL, above sea level; DIAS, Data Integration and Analysis System; EFEO, École française d'Extrême-Orient; GPR, ground penetrating radar; HEC-GeoRAS, Hydrologic Engineering Center: GIS tools for support of HEC-RAS; HEC-RAS, Hydrologic Engineering Center: River Analysis System; HEC-HMS, Hydrologic Engineering Center: Hydrologic Modeling System; MCS, mesoscale convective system; RMSE, root mean square error; SRTM, NASA Shuttle Radar Topography Mission; TRMM, Tropical Rainfall Measuring Mission ⁎ Corresponding author at: 15 Cottenham Ave, Kensington, NSW 2033, Australia. E-mail address: [email protected] (T. Lustig). 1 Present address, 15 Cottenham Ave, Kensington, NSW 2033, Australia. 2 That Yaśovarman's two sons succeeded him may have been contrary to the rules of succession of the “conical clan”, in which all members are ranked hierarchically in terms of nearness of descent from the common ancestor, in this case, Jayavarman II (Vickery, 1986, 108). Under that process, Jayavarman IV might have become king in 910 and not 18 years later. https://doi.org/10.1016/j.jasrep.2017.11.014 Received 29 May 2017; Received in revised form 10 November 2017; Accepted 10 November 2017 2352-409X/ © 2017 Published by Elsevier Ltd. T. Lustig et al. Journal of Archaeological Science: Reports 17 (2018) 195–211 Fig. 1. Koh Ker on the road between Angkor in present-day Cambodia and Wat Phu in southern Laos. Road alignments from Hendrickson (2007). the north of Prasat Thom, he built a 7-km long weir across the original 2. Site investigations Rongea River valley (Evans, 2010–2011), the longest water-manage- ment structure across a river valley in Khmer history (Fig. 2).3 The It is necessary to understand how the embankment was designed, to embankment evidently provided both a level link to the Angkorian understand how soon it overtopped. We will first examine deficiencies highway to Wat Phu in today's southern Laos and created the largest in the design of the embankment and the two outlets, and then assess known Angkorian-era artificial lake (Fig. 3). the risk of overtopping by evaluating the flow of water into and out of Just south of the main spillway are the remains of a small temple, the reservoir. now known as Prasat Boeng Voeng. Its western portal has an inscrip- tion, K. 823, dated 863 śaka or 941 CE (Jacques, 2014, 350).4 The temple has a causeway linking it to the embankment. Our investigations 2.1. Methods were limited because the area is known to have landmines. However, we could see from the lidar that the level of the crest of the middle of The bare-earth digital terrain model (DTM) we used was generated the causeway is lower than the level of the crest of the embankment. from airborne laser scanning (lidar) data acquired in a 2012 aerial This lower level probably results from continual erosion by water campaign. The process used to generate the DTM is described in Evans draining through it to the north (Fig. 4), and it is reasonable to suppose et al. (2013). The elevations given as “above sea level” (ASL) in this that it was originally at the same level as the main embankment. The paper are recorded or inferred from the lidar-derived DTM or from the main embankment would not have been built to match the level of the processed lidar point cloud. According to the data specifications for the temple causeway, but the other way around. On the assumption that the lidar, absolute vertical accuracy is ± 0.15 m RMSE (root mean square temple was built at the same time as its causeway, this implies that the error) although data quality achieved was far superior to this specifi- main embankment was constructed no later than 941 CE. While it is cation in most cases, and relative accuracy of points nearby is in the probable that the work to build the main embankment could have only order of cm-level. been undertaken by a king, further archaeological investigation of the We investigated the spillway (Section 2.5.1) on site by clearing it of causeway is required to help verify that the northern reservoir and grass and shrubs, and surveying the laterite surface with a total station access road were built before 941 CE, presumably by Jayavarman IV. to mm-accuracy. Where we suspected blocks were covered with soil, we But this infrastructure was not to last. Whereas the Rongea River located them with 1.5 cm diameter steel rods hammered into the originally ran eastwards past Point A in Fig. 2, we see that it now runs ground to see if they encountered laterite over an informal grid pattern, northwards, having overtopped and broken through the embankment at and positioned with the total station, noting the depth to refusal. The Point B. Once this happened, the large lake, the unimpeded access road, probes could reach to about 1.8 m below ground. Laterite blocks that and any economic benefits from the reservoir would have been lost. Did were washed downstream were located with a handheld GPS unit to this happen before the political centre moved back to Angkor, and were ~10 m accuracy. Where the blocks were in piles, only the boundaries of the two events linked? the piles were surveyed using the GPS unit. The chute was cleared of grass and shrubs from 10 m upstream of the within-chute structures located on the lidar (Section 2.5.2) down to 3 There were indeed longer Angkorian structures across rivers, such as an embankment the downstream end of the sloping ground, taken to be the toe of the about 30 km long at Angkor, but these were built on the Puok-Siem Reap Delta, where the chute's spillway. Where only a thin layer of soil covered laterite blocks, fl ows were distributed into many channels, the slopes were shallower, and the water this was removed, particularly in the area around the pavement and the flowed more slowly, making it much easier to control.